The difficulty involved in reclaiming portions of a legacy /8 depend on how it has been used. In some cases it could be relatively easy to free up large, contiguous blocks and in other cases the space is highly fragmented.

The technical problems can doubtless be solved with time and money. I think the other issue is the incentive for the user of the address space to go through the renumbering process.

Three RIRs are discussing proposals to introduce mechanisms to allow the transfer of address blocks between their members. (A new proposal was introduced in ARIN). Perhaps those policies (if they are agreed) will provide that incentive?

Leo

I don’t want to go too far with Jay Daley’s argument (I think he’s making more of an equity argument rather than saying that it would meaningfully extend the life of the IPv4 pool), but I have a question about those legacy assignments.

You point out that many of these old /8 networks are at least partially assigned; assuming the allocations were organized enough for this, could the /8 be reclaimed and the assigned parts reassigned as /16s or something else lesser? How much work would that really be?

Regards

Michele

The RIRs were allocated 13 /8s last year. That means that even if the demand for space does not increase, buying one year would involve recovering almost one third of the legacy assigned space from networks that are actively using it.

You seem to be proposing that legacy assignments should be reclaimed even though they are in use. Is that right?

So whilst reclaiming this space might only buy a year or two at the most, it would go a long way towards showing that management of address space by IANA and the RIRs is equitable and inclusive and there is no need for alternate arrangements.

BTW one or two years could actually make a huge difference in the ability of companies to plan the capital expenditure.

IPv6 has far more addresses (about 340 trillion trillion trillion) but it is not backwards compatible with IPv4 and so the transition requires using both IPv4 and IPv6 at the same time. Unfortunately, that is expensive and does not generate any additional income for ISPs and other network operators. And so take-up has been so slow.

When the IANA and RIR IPv4 free pools are empty, ISPs and other network operators will probably find it very expensive to obtain additional IPv4 space and IPv6 will become more attractive. But deploying IPv6 means making changes to a lot of systems. As an example of the sorts of changes that need to happen, when Randy Bush added IPv6 to his network he found that he had problems with DNSBLs not supporting IPv6 addresses. There were problems with other software, too.

There’s a lot of work still to be done.

ICANN staff has developed a way of measuring some (but not all) use of unallocated IPv4 address space. As John wrote a few weeks ago, we can measure the queries received at l.root-servers.net. We have used this to begin begin measuring the problem.

We are continuing to work on this topic and I hope we can write a blog article with some details fairly soon.